Dev Vaghani
Backend and infrastructure engineer · Java, Kubernetes, AWS, Python
Four years on systems where an error costs money the same second it happens.
Bangalore, India · vaghanidev412@gmail.com · vaghani.dev · linkedin.com/in/dev-vaghani · Google Scholar
Independent Work
Downstox: broker-agnostic markets research platform
downstox.com · 2023 - Present- Designed, shipped and operate the whole platform solo: live NSE and BSE ingestion through to screeners, valuation and portfolio tooling. No team, no paywall, no shared on-call.
- 40 tools across screening, DCF valuation with Monte Carlo simulation, and cross-broker portfolio analysis over read-only broker connections, plus PRISM, a signals engine scoring geopolitical events for market impact. Market data archived back to 2016.
- Runs on ₹3,500 (about $37) of infrastructure a month, which constrains every design decision from ingestion to serving.
Experience
SDE-2, Upstox (Indian retail stockbroker)
Bangalore · Jul 2022 - PresentRMS · Post Trade · Transactions · Platform
- Sole owner of the Risk Management System at a broker with 2.4 million active clients, and a contributor to the order management service it sits in front of. Every order passes through it.
- Designed the post-trade square-off path against circuit limits, illiquid scrips and a fixed latency budget, where the failure mode is realised loss rather than a stack trace.
- Migrated a Java service using GenAI tooling, including the verification strategy for deciding whether the output was correct.
Earlier engineering internships
- CleverTap (Jan - Jun 2022). Dashboard features and Keycloak auth. Vue.js
- Altair (Aug - Dec 2021). Build and assembly pipeline for HyperWorks. C++, Python, Tcl/Tk
- Fyle (Feb - Jul 2021). 20+ pages shipped, PageSpeed raised site-wide. Webflow
Publications
- Bandwidth Efficient Digital Image Watermarking Scheme Using a Concatenation of Three Transforms F. N. Thakkar, D. Vaghani, P. Vasoya, D. Desai, 2020
- An Approach for Representation of Node Using Graph Transformer Networks D. Vaghani
Skills
- AI engineering:
- Agent orchestration, Tool calling, Model Context Protocol, LLM APIs (Claude, OpenAI), RAG, Embeddings & vector search, Document AI
- Evals & observability:
- Offline eval sets, LLM-as-judge, Tracing & observability, Latency & token-cost tuning
- Programming:
- TypeScript, Python, Java, SQL, C++
- Backend & platform:
- Node.js, Express, Spring Boot, MongoDB, Socket.io, AWS, Azure, Kubernetes, Docker
- Frontend:
- React, Next.js, Vue, Tailwind
Education
B.E. Information Technology, G H Patel College of Engineering & Technology2018 - 2022